In Defence of Liberty

Driven by data; ridden with liberty.

Statistics and Lampposts XV: Mutation through Resonation

Statistics can be become flawed through indirect communication. (Edited: Stinging Eyes)

Statistics can be become flawed through indirect communication. (Edited: Stinging Eyes)

Caroline Criado-Perez is described as a “freelance journalist and feminist campaigner”, and recently wrote a New Statesman article arguing that domestic violence should be made a separate crime from general violence. In the original version of the article, Ms Criado-Perez asserts:

And make no mistake: it is an epidemic. Domestic abuse is the largest cause of morbidity in women aged 19-44, more than war, cancer or motor vehicle accidents.

Ms Criado-Perez linked to the 2002 World Health Organisation ‘World Report on violence and health’. It stands at 343 pages. This is false attribution, because at no point during the document are people partitioned into the cohorts Ms Criado-Perez describes.

For clarity, mortality means death and morbidity means ill-health. On page 286, the document provides a morbidity ranking by disability-adjusted life years (DALYs). Interpersonal violence is 43rd highest cause of morbidity for women of all ages worldwide, with 0.5% of the total.

This statistic and its variations have been repeated worldwide. A March 2009 edition of the BBC Ten O’Clock News claimed:

The government says incidents of domestic violence have fallen substantially but that it still affects 1 in 4 women. Indeed, for women aged between 15 and 44 it’s the biggest cause of mortality.

A Huffington Post article entitled “50 facts about domestic violence” states:

Order of causes of death for European women ages 16-44: domestic violence, cancer, traffic accidents.

The Huffington Post article cites an Amnesty International document, which in turn cites an English edition of Le Monde diplomatique, which in turn cites the Council of Europe’s Report on Domestic Violence from 2002. The source of the Council of Europe’s claim is unclear:

Statistics shows that for women between 16 and 44 years of age, domestic violence is thought to be major cause of death and invalidity, ahead of cancer, road accidents and even war.

As Ms Criado-Perez’s article highlights, countries may not place domestic violence in a separate legal category. This means that domestic violence figures are usually estimated from crime statistics on general violence.

More or Less

On the BBC’s More or Less in 2009, Tim Harford investigates these claims, and concludes that they are “thankfully, completely wrong”. Looking at the Office for National Statistics figures for England and Wales, Mr Hartford states:

In the year 2007, nearly 6,000 women aged between 15 and 44 died. What you claim is the largest cause of death depends how broad your categories are, but over 2,000 of those deaths were from cancer and other tumours. 1,100 women were killed by ‘external causes’.

It may seem this is a simple confusion between mortality and morbidity, but even the morbidity variant is unsound. Colin Mathers, who heads the World Health Organization on health-related statistics, is invited onto the More or Less programme. Mr Harford asks Dr Mathers if domestic violence could be a leading cause of morbidity, calculated in DALYs:

At a global level, depression is the leading cause of lost healthy life in women aged 15 to 44. That is followed by schizophrenia and bi-polar disorder; so three mental disorders are responsible for the most lost years of healthy life from the non-fatal conditions.

Domestic violence is certainly a significant risk factor – one of the contributors to loss of health – but I doubt very much it will be the leading one.

This pseudo-fact’s mortality variant is noteworthy because of its striking dissonance with the other commonly-cited statistic about domestic violence: “on average, 2 women a week killed by a current or former male partner”. This statistic has the valuable property of being broadly true, if potentially out-dated. In 2002-03, 104 women aged 16 and above were killed by their current or former partner. In the following years, this figure oscillated around 95 women per year, before falling to 76 women in 2012-13 – which is less than 1.5 women per week. For contrast, 15 men were killed by a current or former partner in 2012-13. However, how could the leading cause of mortality possibly contain only around 100 entries?

The rise of flawed statistics demonstrates how information can be corrupted by indirect communication. Media outlets and government departments began citing each other, rather than a central and credible source: this is mutation through resonation.



This entry was posted on August 29, 2014 by in Statistics and tagged , .
%d bloggers like this: